Radar Detection and Waveform Identification under High Interference in CBRS Band

This repository offers a modular framework for radar detection and waveform identification in the Citizens Broadband Radio Service (CBRS) band, targeting environments with significant interference from commercial systems such as 5G. The repository includes two machine learning-based pipelines, each tailored to different input modalities—In-phase/Quadrature (IQ) data and spectrograms.

---

🔹 Introduction

As spectrum sharing becomes increasingly vital for accommodating commercial wireless growth, ensuring reliable coexistence with incumbent users—such as military radar systems—remains a key challenge. This repository addresses that challenge through two machine learning pipelines designed to detect and classify radar signals in shared-spectrum scenarios with high interference.

The framework is adaptable to various signal types and interference conditions, making it suitable for research and prototyping in spectrum sharing, signal classification, and wireless coexistence technologies.

---

🔹 Setup and Prerequisites

System Requirements

• Python 3.8+: Install Python
• Jupyter Notebook: Install Jupyter
• TensorFlow 2.6+: Install TensorFlow

Install additional dependencies listed in the requirements.txt file (if available) using:

🔹 Pipelines

The repository contains two primary pipelines, each supporting both radar detection and waveform classification.

Pipeline 1: IQ-based Radar Detection and Waveform Identification:

This pipeline uses raw IQ samples as input to train and evaluate ML models.

1. Data Preparation: Use the provided MILCOM_2025_Data directory for preprocessed IQ data.
2. Model Training: Open IQ_ML_For_Detection_and_Identification.ipynb to define, train, and validate the model.
3. Model Saving and Inference: Save the trained model and reload it for further evaluation or prediction tasks.
4. IQ_based Waveform Identification To perform waveform classification, adjust the output layer (softmax node) to size 6 and use BIN data from the dataset.

Pipeline 2: Spectrogram-based Radar Detection and Waveform Identification:

This pipeline operates on spectrogram representations of radar signals

1. Preparing Your Data:Load spectrogram data from the MILCOM_2025_Data directory.
2. Run the Script:Run the notebook ML_For_Spec_Radar_Detection_ViT.ipynb directly without modifications.
3. Model Saving and Inference: Save the trained model for reuse and load it during inference or evaluation.
4. Spectogram_based Waveform Identification Use the notebook Spec_ML_For_Identification.ipynb and the provided BIN data files for classification tasks.

🔹 Running the Notebooks

• After setting up either pipeline:
• Launch Jupyter and open the respective notebook.
• Follow the inline instructions to train, save, and evaluate the model.
• Use the trained models to predict or classify new radar signals under interference conditions.

🔹 Citation

This work is submitted to MILCOM 2025.

Shafi Ullah Khan, Michel Kulhandjian, and Debashri Roy,“Pushing the Boundaries in CBRS Band: Robust Radar Detection within High 5G Interference,” In IEEE Military Communications Conference (MILCOM), October 2025 [Accepted].